Make Intel great again? Intel i7-11700 Rocket lake processor review!

[cagoblex]

Hello everyone. After delays, delays and delays I finally got my package from DHL. And with the package, I got a motherboard, which I’m not allowed to show you right now, and two processors, you can take a guess what they are. And, two more processors, which I will test in this review. So it’s a 11th generation Rocket lake i7 11700 and i9 11900T. I will post two separate reviews for these two as they are intended for different audiences. In this review, we will take a close look at the i7 11700.

I have reviewed a ES i7 11700 before with the Dell system, but that ES is only 1.8GHz in frequency and is one stepping before retail. Today however, we will be looking at the actual 11700 CPU. It is technically a sample, but it’s a Qualification sample or QS, which is usually sent out to media for review or to manufacturers for demo systems. 99% of times they are the same as the retail chip, I think the only exception I can think of is Sandy bridge. It was 10 years ago and I couldn’t remember what exactly was different. But the difference is very minor. So it’s safe to say what we are looking at today is what you will get if you buy an i7 11700 from your favorite retailer.

Enough said, let’s take a look it at the processor itself. Of course this review is before the media embargo for Rocket lake, so Intel will not share any detail information with me on this. Even if I do know some, I cannot share it right now. But I will use public information and leaks online, and that pretty much tells all you need to know about Rocket lake.
 

First of all, this is a slide Intel showed at CES2021, and here are the key differences.

Unlike Comet lake, the core count for flagship model has reduced from 10 cores 20 threads to 8 cores 16 threads. This is due to the addition of AVX512 and the change in architecture. But it brings higher IPC and better cache performance compares to Comet lake.

It supports DDR4 3200 memory, which is not a big deal for computer enthusiasts since you will be overclocking your memories anyways. But it matters for some prebuilds like Dell and HP systems which does not give you the option for XMP and memory overclocking.

It employs the new Intel XE graphics architecture, and supports AV1 decoding and HDMI2.0b output. It has 32 EU and performs much better than UHD630.

It adds support for PCI-E4.0 (finally!) and provides 20 lanes. Which means you can connect both your GPU and SSD directly to CPU.

The new Z590 chipset it pairs with supports USB3.2Gen2x2 20Gbps and x8 DMI bus. Which doubles the speed of Z490, giving more bandwidth to high speed IOs and PCH attached M.2 SSDs.



Things looks pretty good on paper here. But would Intel actual deliver what they promised? Let’s find out!



First of all, for this generation, i7 and i9 will have identical core and thread count, as well as cache speed and size. The only difference is that i9 supports TVB, or Turbo Velocity Boost. The i9 11900K actually has 100MHz lower base clock than i7 11700K, which is quite strange in terms of product positioning.
 

 

For the CPU we are looking at today, it’s a non K version of the i7 11700. It has 2.5GHz base and 4.9GHz boost clock. Which is much higher than the sample I reviewed last time. Let’s get started.


CPU Performance
 

For the review today we are using the Asus Maximus XIII Hero we reviewed last time. For the memories I am using a pair of G Skill Trident 3200MHz CL14 memories with XMP1 profile. For storage I am using a Western Digital SN850 1TB PCI-E4.0 M.2 SSD. For video card I am using a MSI Gaming X Trio RTX 3080 with default BIOS. I have included Ryzen 7 5800X for comparison. It is using the Asus X570 TUF motherboard with the same memories and SSD under the same settings. PBO is set to auto in BIOS.



First is the CPU-Z benchmark.
 

In this test, thanks to the increase in frequency, the 11700 is achieving a much higher single thread score compares to our 1.8Ghz sample, and it’s about 14% increase higher than the i7 10700. It also managed to achieve about 3.5% higher score than Ryzen 5800X. In multi core testing, it scales quite well compares to single core test. It is about 8% faster than 10700 and it also beats the 5800X by a tiny bit. Please note that this is a non-K i7, it will be priced in the upper $300 range, which is about $80 cheaper than the 5800X. So the result looks pretty promising to me.

Next is the Cinebench R15 and R20.
 

For Cinebench R15, we tested the single core, multi core, as well as OpenGL with the integrated iGPU. The 11700 performs 17% higher single core score, which is very close to what Intel claims. It still lags behind the 5800X in both single and multi core scores.



However the iGPU bring a very decent increase. It has a 30% increase in average FPS in OpenGL test compares to it’s predecessor.
 

 

In Cinebench R20, it is still about 10% slower than 5800X in multi core performance, and only about 2% slower than 5800X in single core performance. It also brings 2 digit IPC increase over the 10700, which again matches what Intel claims.



Next is the Aida64 memory and cache benchmark. As I mentioned couple times in the previous reviews, Rocket lake brings a decent increase in cache performance, and the test result confirms that. I didn’t include the 5800X as they have different architecture and the numbers should not be compared directly. We will just take a look at how much performance increase it brings compares to 10700.
 

 

For L1 cache, we are seeing more than double the performance in all copy, write and read benchmarks on our 11700 compares to 10700. This contributes a lot to the IPC increase.

 

 

 

The bump in L2 cache is not as dramatic as the L1 cache, but it is still quite impressive. We are seeing a 50% increase in speed on read and copy, and the write speed remains similar.
 

 

The 11700 loses some points here in L3 cache read performance, but remains relatively the same in write and copy.
 

 

The memory part is a little interesting. It achieves higher copy and write, but lags behind in read. I was informed by the product manager of Z590 boards at Supermicro that Intel has released a microcode patch to fix the problem. We will have to wait till the motherboard manufacturers release new BIOS updates to see if it will fix the problem.
Here is a screenshot of the email.

 

Next will be 7 zip.
 

 

To my surprise, 5800X performs really well here and out performs the 11700 by a significant margin. However compares to 10700, the 11700 is also able to achieve about 10% increase in performance, which is decent for a ‘tock’ product.

 

The Y-cruncher results are pretty confusing. 11700 is able to finish calculating 1 billion digits of PI in just 177 seconds, which is almost half the time of the 10700 and about 50% faster than 5800X. However it lags behind the 5800X in the single core benchmark, actually it even lags behind 10700 on single core. Y-cruncher actually supports AVX512 which would give Rocket lake a significant boost in performance. So the high multi core score is easy to explain. However for single core, I guess AVX512 wasn’t properly utilized, which results to the significant hit in performance. I am still looking into this and will update the review once I find out.
 

In Blender, we rendered both BMW and classroom. The 11700 is again much faster than our 1.8Ghz sample, and catches close to 5800X.
 

 

 

In V-ray, I recorded the average frequency during the benchmark. The i7 11700 averages 200Mhz lower than the 10700, while still out performs by 12%. It is still 10% slower than the 5800X.
 

 

 

In handbrake, we are transcoding a 1:31 4K video into H.264 1080P 30FPS. It’s a tie between 5800X and 11700. They are both 30% faster than the 10700. Good job Intel.
 

 

Lastly for the CPU part, is Geekbench. It outperforms the 5800X in single threaded testings by a tiny margin, while being 30% faster than the 10700. It lags behind 5800X in multi core testing by 5%, which being 10% faster than 10700.
 

iGPU part- Intel UHD750 Aka XE32



Next will be the GPU part. Let’s first start by the iGPU. Intel has finally push it’s XE graphics to the desktop market. This GPU is called UHD750 and has 32 XE architecture EUs. I happened to have it’s great great ancestor in my collection, this is a early GPU prototype right before Larabee. You are getting old if you know what I’m talking about lol. So this GPU is benchmarked again with the latest driver from Intel and let’s see how it performs.

 

View attachment intel.webp
 

 

 

 

In Geekbench 5 Compute test, it delivers very promising results. It is 15% faster in Vulkan and 20% faster in OpenCL. This is GTX650Ti level of performance. I know it may not sounds very fast but considering it’s tiny size, and ultra low TDP, it is still a big step in iGPU. Not to say you get it for free anyways.  Not very impressed? Let’s move on.

Impressive? The 3Dmark score shows a 47% performance increase over the UHD630 in 10700! Again, Intel actually delivers what they promised this time.



Gaming part- With MSI X Trio RTX 3080

Next let’s see how it performs with the a discrete GPU. The GPU we are using is a MSI Gaming X Trio RTX 3080. I’m using the default BIOS for all benchmarks.
 

Firstly is 3Dmark Time Spy.

I used smaller scales in the chart just to show the difference. But if you actually do the math, it is only about 8% slower than the 5800X, and 4% faster than it’s predecessor.
 

In hitman2, the result is similar to Time spy, only 5800X is about 25% faster this time, instead of 10%.
 

 

In Horizon it’s a different story. It performs on par with the 5800X, while being about 8% faster in GPU FPS. The only reason I can think of is the benefit of PCI-E4.0.
 

 

 

In dirt 5, again I changed the scale to show the difference. Because I think it would be boring if they all look the same on the chart. The actual difference is less than 5% between the 5800X and 11700, but it’s still a win.
 

 

 

In shadow of the Tomb Raider, 5800X also pulls ahead in FPS, but 11700 is about 10% faster in GPU FPS. However that did not translate to actual FPS, the FPS is determined by the lowest of the three. It is about 22FPS less than 5800X and 13FPS faster than 10700.


Next let’s take a look at Unigine tests.
 

 

In superposition, all three processor performs very similar. It is more sensitive to frequency rather than architecture.
 

For Unigine heaven, the trend continues. This is a rather old benchmark, so it’s not very processor intensive. All processors performs similar and the difference is within 5%.
 

 

For Unigine Valley there is a less than 10% increase compares to 10700. It beats the 5800X, but less than 1%.
 

So conclusion. Did Intel deliver what they promised? The answer is a clear yes. But does this mean you should spend your hard earned dollars and upgrade to a Rocket lake right now? Well it depends. If you are upgrading from a Z490 it’s probably not worth it in my opinion. But if you are upgrading from something like Skylake and you just want a good gaming PC, then I would say it’s definitely worth the upgrade. Yes I know Alder lake is coming in a couple months, but the cost of the platform would be much higher because of DDR5 and new power delivery requirements for the CPU.

Not to say, if it’s really good like the Ryzen 5000 series, most likely you won’t be able to get one at MSRP. We will have another review of the 35W i9 10900T soon. So please stay tuned.

You can watch a video version of the review here:



Thanks!

[porina]

On y-cruncher result, ram performance may be a contributor. It uses a fair chunk of ram, around 4.7GB at that test size. The writer of the software also said you can approximate it as 50% each of read/writes, so a reduction in one will impact overall performance. Single thread doesn't make a high demand so wouldn't be affected. Multi-thread will stress the ram. It would be interesting if you get the microcode update and try that again.

 

It also reminds me there may be some problem with it running on Zen 3 with SMT on in that SMT off results may be higher, which is not expected.

That's writer of y-cruncher commenting on Anandtech's results when Zen 3 came out. I'm not aware of any further activity since then.

 

Still that 5800X time of 41 seconds equals my own testing on a 3700X at 4.2 GHz with 3600C16 ram.

 

I think on the past thread with the ES I requested a Prime95 bench, is that at all possible? Would give an additional insight to what AVX-512 is doing.

[cagoblex]

6 minutes ago, porina said:

On y-cruncher result, ram performance may be a contributor. It uses a fair chunk of ram, around 4.7GB at that test size. The writer of the software also said you can approximate it as 50% each of read/writes, so a reduction in one will impact overall performance. Single thread doesn't make a high demand so wouldn't be affected. Multi-thread will stress the ram. It would be interesting if you get the microcode update and try that again.

 

It also reminds me there may be some problem with it running on Zen 3 with SMT on in that SMT off results may be higher, which is not expected.

That's writer of y-cruncher commenting on Anandtech's results when Zen 3 came out. I'm not aware of any further activity since then.

 

Still that 5800X time of 41 seconds equals my own testing on a 3700X at 4.2 GHz with 3600C16 ram.

 

I think on the past thread with the ES I requested a Prime95 bench, is that at all possible? Would give an additional insight to what AVX-512 is doing.

Thanks for the info. That really helps. Actually I did the prime 95 test but did not include it in this review. I plan to add it to the i9 11900T review as I think power consumption is more of a concern for low power chips. But yes there is a big difference in power consumption with AVX512 on/off and it affects the frequency behavior as well. I’ve been really busy with work recently but I will try to publish the 11900T review this week. 

[Letgomyleghoe]

all that data is so weird and back and worth, intel's finally made a worthy competitor if they can keep the prices cheaper. 

[cagoblex]

2 hours ago, Letgomyleghoe said:

all that data is so weird and back and worth, intel's finally made a worthy competitor if they can keep the prices cheaper. 

The prices are pretty competitive this time. But the launch has been delayed to April. Just got notified by one of major OEMs.